Devstral 2
Mistral · 1 configuration
Intelligence8.6
Input / 1M$0.40
Output / 1M$2.00
Context256K
Tariff: Mistral · Source: models-dev · USD per million tokens
Thinking configuration
What does more thinking buy you?
Compare this model’s measured thinking configurations.
Scroll the chart ↔ · tap a point for details
Efficient frontier: best measured score at each cost1 measured configurations · 0 without both values
Selected: thinking default · Click a point to switch configuration
Mistral's coding-agent model for repository work, terminal tasks, and software fixes
Intelligence IndexArtificial Analysis composite index, not a percentage. Different versions of the index are not directly comparable.
8.6 pointsGPQA DiamondGraduate-level scientific questions. Measures specialist reasoning rather than web-search quality.
59.4 %Humanity’s Last ExamDifficult questions across many disciplines, compared under the same evaluation setup.
3.6 %SciCodeCode generation for scientific research problems.
32.8 %Terminal-Bench 2.1Agent tasks in a terminal environment. Results also depend on the evaluation harness.
30.3 %τ²-BenchTool use and interaction in assistance scenarios. Does not replace an evaluation of your integration.
24.9 %IFBenchVerifiable instruction following. A proxy for precision rather than a benchmark of creative style.
38.1 %Omniscience accuracyFactual accuracy in Artificial Analysis's Omniscience evaluation.
20.8 %Specifications from the catalog
Inputtext
Outputtext
Tool callingYes
Structured outputNot available
SpeedNot available
First tokenNot available
Maximum output262.1K
Catalog listing date2025-12-09